Re: [PATCHv4 8/8] Convert remaining complex function to generated files

["Paul E. Murphy" <murphyp at linux dot vnet dot ibm dot com>]

On 08/26/2016 04:22 PM, Joseph Myers wrote:
> On Fri, 19 Aug 2016, Paul E. Murphy wrote:
> 
>> On 08/19/2016 03:45 PM, Joseph Myers wrote:
>>> On Thu, 18 Aug 2016, Paul E. Murphy wrote:
>>>
>>>> +#if M_LIBM_NEED_COMPAT (clog10)
>>>> +  /* __clog10 was accidentally exported in past versions.  */
>>>
>>> I'm not sure that comment is accurate.  __clog10 is used in tgmath.h to 
>>> provide support for type-generic log10 for complex arguments.  That looks 
>>> fairly deliberate to me as a public ABI.  One could argue that tgmath.h 
>>> should actually use clog10 not __clog10 there, since it's inside a 
>>> conditional on __USE_GNU, but one could also argue that __clog10 should be 
>>> exported for use by the C++ standard library which has to provide log10 
>>> for complex numbers (libstdc++ doesn't currently use clog10 or __clog10 
>>> for this, but could, and logically it should be __clog10 for namespace 
>>> reasons; cf past discussions about providing more __* exports to eliminate 
>>> the need for g++ to define _GNU_SOURCE by default).
>>>
>>
>> How about changing it to:
>>
>> +  /* __clog10 is also a public symbol.  */
> 
> Patches 7 and 8 are OK with that wording and with the wording in 
> sysdeps/ieee754/ldbl-opt/s_clog10l.c
> 
>> +/* The aliasing for long double is a bit more complicated
>> +   since some private symbols leaked out in the past.  */
> 
> also fixed similarly since this may not be a leakage of private symbols.
>

Comments in both math/s_clog10_template.c and ldbl-opt/s_clog10l.c
rewritten as such.

Committed as 1dbc54f and feb62dd.

Thanks for your patience and detail reviewing these.